Tissue-characterization of the prostate using radio frequency ultrasonic signals

In this paper, we will present a complete method and system for the detecti on of prostatic carcinoma, providing color-coded images of the estimated pr obability of malignancy by processing radio-frequency ultrasonic echo signa ls. For this, st hardware setup based on a conventional diagnostic sonograp h was realized. The image-processing software works on ultrasound images au tomatically segmented into regions of about 3 X 3.5 mm. System-dependent ef fects, as well as tissue attenuation, were measured and compensated for. Ti ssue-characterization parameters, which have been used successfully by othe r authors, were calculated for each segment. To demonstrate the methods of selection of relevant parameters and comparison of different classifiers, a first clinical study using data of 33 patients with local prostatic carcin oma was performed. For these patients, location and extent of the carcinoma were known from histological findings after radical prostatectomy. Classif iers investigated during the study were: the linear and quadratic Bayes cla ssifier, a nearest neighbor classifier, and several classifiers based on Ko honen-maps. The best classifier was used to calculate color-coded result im ages. Applying a threshold of 50% to the estimated probability of malignanc y, produced the encouraging results of 82 and 88% for sensitivity and speci ficity, respectively.